Pin the Tail on the PY Donkey

   The winter’s drawing in, so plenty of time for theorising!

The picture above is a hypothetical plot of personal PYs assigned to every sailor in a class.  There are 400 sailors in the class and 133 of them have sailed to a PY of 1065 (as shown by point B), 60 sailors have achieved a PY of 1085 (point C) etc.

When working out the PY of the class as a whole, what number would you pick?

A (1000) is the PY of the fastest sailor.  Chose this is you want the best-in-class to have a fair chance of winning handicap events.

B (1065) is most common PY (the mode).  Chose this if you want the PY to be fair to the most number of people.

C (1085) is PY of the sailor who finishes exactly mid fleet (the median).  Chose this if you think the mid-fleet guy should have a fair chance of beating half of another class.

D (1090) is the PY of the average sailor (the mean).  Chose this if you think the class as a whole should have a fair chance of beating another whole class (if you added scores up team racing style).

E – some other number?

Poll Question: What PY should be used for this class?
Poll Choice	Votes	Poll Statistics
A	12	[34.29%]
B	15	[42.86%]
C	3	[8.57%]
D	1	[2.86%]
Other	4	[11.43%]
You can not vote in this poll

Author	Message Share Topic Printable Version Delicious Digg Facebook Furl Google MySpace Newsvine reddit StumbleUpon Translate Twitter Windows Live Yahoo Bookmarks Topic Search Topic Options Post Reply Create New Topic
Guests Send Private Message Find Member's Posts Guest Group	Post Options Post Reply Quote Guests Quote Reply Topic: Pin the Tail on the PY Donkey Posted: 03 Nov 14 at 9:29pm
	The winter’s drawing in, so plenty of time for theorising! The picture above is a hypothetical plot of personal PYs assigned to every sailor in a class. There are 400 sailors in the class and 133 of them have sailed to a PY of 1065 (as shown by point B), 60 sailors have achieved a PY of 1085 (point C) etc. When working out the PY of the class as a whole, what number would you pick? A (1000) is the PY of the fastest sailor. Chose this is you want the best-in-class to have a fair chance of winning handicap events. B (1065) is most common PY (the mode). Chose this if you want the PY to be fair to the most number of people. C (1085) is PY of the sailor who finishes exactly mid fleet (the median). Chose this if you think the mid-fleet guy should have a fair chance of beating half of another class. D (1090) is the PY of the average sailor (the mean). Chose this if you think the class as a whole should have a fair chance of beating another whole class (if you added scores up team racing style). E – some other number?

Daniel Holman Send Private Message Find Member's Posts Really should get out more Joined: 17 Nov 08 Online Status: Offline Posts: 997	Post Options Post Reply Quote Daniel Holman Report Post Quote Reply Posted: 03 Nov 14 at 9:35pm
	It doesn't matter a jot as long as every other class is handicapped the same way. Only trouble then is that is the 50th percentile or ultimate 1st percentile sailor in Class X as good as the 50th percentile or ultimate 1st percentile sailor in Class Z. Or is the gulf in talent between the 1st and 50th percentile larger in one class than another (this would be a better bet for a pot hunter) Dan

Guests Send Private Message Find Member's Posts Guest Group	Post Options Post Reply Quote Guests Quote Reply Posted: 03 Nov 14 at 9:41pm
	Exactly Dan, different classes will have different skill distributions so you can't give the best guy the same chance of winning as the mid fleet guy has of coming in the middle. So there needs to be a choice on which one the PY system is aimed at - potential winners, the bulk of the fleet, Mr Average, or who?

Daniel Holman Send Private Message Find Member's Posts Really should get out more Joined: 17 Nov 08 Online Status: Offline Posts: 997	Post Options Post Reply Quote Daniel Holman Report Post Quote Reply Posted: 03 Nov 14 at 9:46pm
	To be honest you'd have most luck doing it with the champions. Whilst they aren't created equal, I suspect it'd be closer to achieving parity than using mr Average. Just tough luck if you have a sailing god in your fleet, and don't expect to win the Bloody Mary in an olympic class unless you are literally a gold medallist.

Guests Send Private Message Find Member's Posts Guest Group	Post Options Post Reply Quote Guests Quote Reply Posted: 03 Nov 14 at 9:51pm
	Ah, but the Olympians don't handicap race, so never influence the numbers! By the way, you're supposed to press button A if that's your (much respected) opinion.

craiggo Send Private Message Find Member's Posts Really should get out more Joined: 01 Apr 04 Location: United Kingdom Online Status: Offline Posts: 1810	Post Options Post Reply Quote craiggo Report Post Quote Reply Posted: 03 Nov 14 at 10:15pm
	The problem is that while point A has undoubtedly been achieved, how confident are you that external factors weren't involved, like he was the only guy who was on a sheltered part of the course when a squall wiped the rest out, or he was the only one to make it round the top mark before the tide really kicked in preventing the rest from getting around? We all know that different classes have a different average standard and therefore taking point B simply sets the bar at the average for that particular class but doesn't really allow a boat for boat comparison. Given that a measurement system is likely to be completely unworkable, I'd suggest picking a point somewhere between point A and B. For the more competitive high standard fleets the difference between A and B will be relatively small, however if you look at a less competitive fleet, it will represent a bigger shift. I imagine that the removal of outliers in the PYS system gets somewhere towards this assuming they remove more from the back end, so on that basis, I think we probably already have A-/B+ in use and that's where it should be, although my knowledge of the inner workings of PYS is not complete so there is an error band which may lean towards point B.

Old Timer Send Private Message Find Member's Posts Far too distracted from work Joined: 05 Jun 13 Online Status: Offline Posts: 370	Post Options Post Reply Quote Old Timer Report Post Quote Reply Posted: 03 Nov 14 at 10:17pm
	You need an option that says any or all of the above.

Medway Maniac Send Private Message Find Member's Posts Visit Member's Homepage Really should get out more Joined: 13 May 05 Location: United Kingdom Online Status: Offline Posts: 2788	Post Options Post Reply Quote Medway Maniac Report Post Quote Reply Posted: 03 Nov 14 at 10:18pm
	Wasn't this what we touched on when discussing the respective Solo and OK PYs? It was suggested that the spread was appreciably smaller in the OK than the Solo, so that a top sailor in a new boat would be better in off a Solo than in an OK, given that PY currently tends to represent the average. I suggested that if one wanted the boat's potential (i.e. take our the skill factor), one could better take, say, the 90th percentile of the data collected by the RYA rather than the mean, but was told by 'one who seemed to know' that there are 'difficulties' with this approach.
	Wilsonian SC 3000 Class

Guests Send Private Message Find Member's Posts Guest Group	Post Options Post Reply Quote Guests Quote Reply Posted: 03 Nov 14 at 10:33pm
	Craiggo, you can't establish these numbers on one race, so hopefully those factors get evened out over a year or so (but small samples ARE a problem with a purely statistical approach like the PY scheme). I agree though that somwhere between A and B could be best overall. Old timer, sadly it isn't possible to select one number that is equally fair to all groups. I could have 'don't care' as an option, but am figuring those that don't care will just ignore he thread! I thought I read on here recently that the mean is used, but only the mean of data that is not more than 5% slower than the overall mean. Or did I imagine that? One thing I think is interesting is that with a distribution like the one shown, most people are faster than average.

RS400atC Send Private Message Find Member's Posts Really should get out more Joined: 04 Dec 08 Online Status: Offline Posts: 3011	Post Options Post Reply Quote RS400atC Report Post Quote Reply Posted: 03 Nov 14 at 10:55pm
	You just think yourself lucky there are 400 active racers in the class, and try to find a few to have a nice race with.